Adding interconnect properties to the consumer nodes creates
a dependency on noc device. The imx-bus devfreq driver is not usable
without the full interconnect support. The interconnect is not yet
working on i.MX platforms. The devlink created on device_add makes
the lcdif and other nodes that have the interconnect properties
wait for the noc (imx-bus driver) to probe first.
To make sure the interconnect consumers (nodes that have interconnect
properties already added) will still probe, lets disable the noc node
for now. Once the interconnect on i.MX platforms is fully functional,
the status of the noc node can be changed.
Fixes: ad1abc8a03fdbc05b ("arm64: dts: imx8mq: Add interconnect for lcdif")
Signed-off-by: Abel Vesa <[email protected]>
---
arch/arm64/boot/dts/freescale/imx8mq.dtsi | 1 +
1 file changed, 1 insertion(+)
diff --git a/arch/arm64/boot/dts/freescale/imx8mq.dtsi b/arch/arm64/boot/dts/freescale/imx8mq.dtsi
index 972766b67a15..f3182878f596 100644
--- a/arch/arm64/boot/dts/freescale/imx8mq.dtsi
+++ b/arch/arm64/boot/dts/freescale/imx8mq.dtsi
@@ -1305,6 +1305,7 @@ noc: interconnect@32700000 {
fsl,ddrc = <&ddrc>;
#interconnect-cells = <1>;
operating-points-v2 = <&noc_opp_table>;
+ status = "disabled";
noc_opp_table: opp-table {
compatible = "operating-points-v2";
--
2.31.1
On Thu, Nov 11, 2021 at 01:16:09PM +0200, Abel Vesa wrote:
> Adding interconnect properties to the consumer nodes creates
> a dependency on noc device. The imx-bus devfreq driver is not usable
> without the full interconnect support. The interconnect is not yet
> working on i.MX platforms. The devlink created on device_add makes
> the lcdif and other nodes that have the interconnect properties
> wait for the noc (imx-bus driver) to probe first.
>
> To make sure the interconnect consumers (nodes that have interconnect
> properties already added) will still probe, lets disable the noc node
> for now. Once the interconnect on i.MX platforms is fully functional,
> the status of the noc node can be changed.
>
> Fixes: ad1abc8a03fdbc05b ("arm64: dts: imx8mq: Add interconnect for lcdif")
Martin,
Do you have any comment? So your commit added something untested?
Shawn
> Signed-off-by: Abel Vesa <[email protected]>
> ---
> arch/arm64/boot/dts/freescale/imx8mq.dtsi | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/arch/arm64/boot/dts/freescale/imx8mq.dtsi b/arch/arm64/boot/dts/freescale/imx8mq.dtsi
> index 972766b67a15..f3182878f596 100644
> --- a/arch/arm64/boot/dts/freescale/imx8mq.dtsi
> +++ b/arch/arm64/boot/dts/freescale/imx8mq.dtsi
> @@ -1305,6 +1305,7 @@ noc: interconnect@32700000 {
> fsl,ddrc = <&ddrc>;
> #interconnect-cells = <1>;
> operating-points-v2 = <&noc_opp_table>;
> + status = "disabled";
>
> noc_opp_table: opp-table {
> compatible = "operating-points-v2";
> --
> 2.31.1
>
Am Dienstag, dem 23.11.2021 um 16:58 +0800 schrieb Shawn Guo:
> On Thu, Nov 11, 2021 at 01:16:09PM +0200, Abel Vesa wrote:
> > Adding interconnect properties to the consumer nodes creates
> > a dependency on noc device. The imx-bus devfreq driver is not
> > usable
> > without the full interconnect support. The interconnect is not yet
> > working on i.MX platforms. The devlink created on device_add makes
> > the lcdif and other nodes that have the interconnect properties
> > wait for the noc (imx-bus driver) to probe first.
> >
> > To make sure the interconnect consumers (nodes that have
> > interconnect
> > properties already added) will still probe, lets disable the noc
> > node
> > for now. Once the interconnect on i.MX platforms is fully
> > functional,
> > the status of the noc node can be changed.
> >
> > Fixes: ad1abc8a03fdbc05b ("arm64: dts: imx8mq: Add interconnect for
> > lcdif")
>
> Martin,
>
> Do you have any comment? So your commit added something untested?
>
> Shawn
hi Shawn,
well, for imx8mq the only missing piece is the mxsfb icc bandwidth
request. I posted a first version a year ago but that didn't make it
in:
https://lore.kernel.org/linux-arm-kernel/[email protected]/
So this should create a working state until the real fix in mxsfb is
there (although I'd revert commit ad1abc8a03fd ("arm64: dts: imx8mq:
Add interconnect for lcdif") instead). I am indeed to blame for this
and I'm sorry I haven't yet taken the time to work on mxsfb again.
I can say I plan to do so soon, but without promise that I quickly find
a solution that'll be merged...
If you merge this (or revert ad1abc8a03fd), I'll remember to revert
with said update.
thanks,
martin
>
> > Signed-off-by: Abel Vesa <[email protected]>
> > ---
> > arch/arm64/boot/dts/freescale/imx8mq.dtsi | 1 +
> > 1 file changed, 1 insertion(+)
> >
> > diff --git a/arch/arm64/boot/dts/freescale/imx8mq.dtsi
> > b/arch/arm64/boot/dts/freescale/imx8mq.dtsi
> > index 972766b67a15..f3182878f596 100644
> > --- a/arch/arm64/boot/dts/freescale/imx8mq.dtsi
> > +++ b/arch/arm64/boot/dts/freescale/imx8mq.dtsi
> > @@ -1305,6 +1305,7 @@ noc: interconnect@32700000 {
> > fsl,ddrc = <&ddrc>;
> > #interconnect-cells = <1>;
> > operating-points-v2 = <&noc_opp_table>;
> > + status = "disabled";
> >
> > noc_opp_table: opp-table {
> > compatible = "operating-points-v2";
> > --
> > 2.31.1
> >
On Tue, Nov 23, 2021 at 09:15:57AM +0000, Martin Kepplinger wrote:
> Am Dienstag, dem 23.11.2021 um 16:58 +0800 schrieb Shawn Guo:
> > On Thu, Nov 11, 2021 at 01:16:09PM +0200, Abel Vesa wrote:
> > > Adding interconnect properties to the consumer nodes creates
> > > a dependency on noc device. The imx-bus devfreq driver is not
> > > usable
> > > without the full interconnect support. The interconnect is not yet
> > > working on i.MX platforms. The devlink created on device_add makes
> > > the lcdif and other nodes that have the interconnect properties
> > > wait for the noc (imx-bus driver) to probe first.
> > >
> > > To make sure the interconnect consumers (nodes that have
> > > interconnect
> > > properties already added) will still probe, lets disable the noc
> > > node
> > > for now. Once the interconnect on i.MX platforms is fully
> > > functional,
> > > the status of the noc node can be changed.
> > >
> > > Fixes: ad1abc8a03fdbc05b ("arm64: dts: imx8mq: Add interconnect for
> > > lcdif")
> >
> > Martin,
> >
> > Do you have any comment?? So your commit added something untested?
> >
> > Shawn
>
> hi Shawn,
>
> well, for imx8mq the only missing piece is the mxsfb icc bandwidth
> request. I posted a first version a year ago but that didn't make it
> in:
> https://lore.kernel.org/linux-arm-kernel/[email protected]/
>
> So this should create a working state until the real fix in mxsfb is
> there (although I'd revert commit ad1abc8a03fd ("arm64: dts: imx8mq:
> Add interconnect for lcdif") instead).
Besides lcdif, the mipi-csi devices have interconnects property too. Are
they already working? If so, it makes more sense to revert ad1abc8a03fd
instead.
Shawn
> I am indeed to blame for this
> and I'm sorry I haven't yet taken the time to work on mxsfb again.
>
> I can say I plan to do so soon, but without promise that I quickly find
> a solution that'll be merged...
>
> If you merge this (or revert ad1abc8a03fd), I'll remember to revert
> with said update.
Am Dienstag, dem 23.11.2021 um 17:24 +0800 schrieb Shawn Guo:
> On Tue, Nov 23, 2021 at 09:15:57AM +0000, Martin Kepplinger wrote:
> > Am Dienstag, dem 23.11.2021 um 16:58 +0800 schrieb Shawn Guo:
> > > On Thu, Nov 11, 2021 at 01:16:09PM +0200, Abel Vesa wrote:
> > > > Adding interconnect properties to the consumer nodes creates
> > > > a dependency on noc device. The imx-bus devfreq driver is not
> > > > usable
> > > > without the full interconnect support. The interconnect is not
> > > > yet
> > > > working on i.MX platforms. The devlink created on device_add
> > > > makes
> > > > the lcdif and other nodes that have the interconnect properties
> > > > wait for the noc (imx-bus driver) to probe first.
> > > >
> > > > To make sure the interconnect consumers (nodes that have
> > > > interconnect
> > > > properties already added) will still probe, lets disable the
> > > > noc
> > > > node
> > > > for now. Once the interconnect on i.MX platforms is fully
> > > > functional,
> > > > the status of the noc node can be changed.
> > > >
> > > > Fixes: ad1abc8a03fdbc05b ("arm64: dts: imx8mq: Add interconnect
> > > > for
> > > > lcdif")
> > >
> > > Martin,
> > >
> > > Do you have any comment? So your commit added something
> > > untested?
> > >
> > > Shawn
> >
> > hi Shawn,
> >
> > well, for imx8mq the only missing piece is the mxsfb icc bandwidth
> > request. I posted a first version a year ago but that didn't make
> > it
> > in:
> > https://lore.kernel.org/linux-arm-kernel/[email protected]/
> >
> > So this should create a working state until the real fix in mxsfb
> > is
> > there (although I'd revert commit ad1abc8a03fd ("arm64: dts:
> > imx8mq:
> > Add interconnect for lcdif") instead).
>
> Besides lcdif, the mipi-csi devices have interconnects property too.
> Are
> they already working? If so, it makes more sense to revert
> ad1abc8a03fd
> instead.
imx8mq-mipi-csi.c request the bandwidth, yes. we use that and the
preliminary mxsfb request above so that works.
do you want me to send the revert?
martin
>
> Shawn
>
> > I am indeed to blame for this
> > and I'm sorry I haven't yet taken the time to work on mxsfb again.
> >
> > I can say I plan to do so soon, but without promise that I quickly
> > find
> > a solution that'll be merged...
> >
> > If you merge this (or revert ad1abc8a03fd), I'll remember to revert
> > with said update.
On Tue, Nov 23, 2021 at 09:30:02AM +0000, Martin Kepplinger wrote:
> Am Dienstag, dem 23.11.2021 um 17:24 +0800 schrieb Shawn Guo:
> > On Tue, Nov 23, 2021 at 09:15:57AM +0000, Martin Kepplinger wrote:
> > > Am Dienstag, dem 23.11.2021 um 16:58 +0800 schrieb Shawn Guo:
> > > > On Thu, Nov 11, 2021 at 01:16:09PM +0200, Abel Vesa wrote:
> > > > > Adding interconnect properties to the consumer nodes creates
> > > > > a dependency on noc device. The imx-bus devfreq driver is not
> > > > > usable
> > > > > without the full interconnect support. The interconnect is not
> > > > > yet
> > > > > working on i.MX platforms. The devlink created on device_add
> > > > > makes
> > > > > the lcdif and other nodes that have the interconnect properties
> > > > > wait for the noc (imx-bus driver) to probe first.
> > > > >
> > > > > To make sure the interconnect consumers (nodes that have
> > > > > interconnect
> > > > > properties already added) will still probe, lets disable the
> > > > > noc
> > > > > node
> > > > > for now. Once the interconnect on i.MX platforms is fully
> > > > > functional,
> > > > > the status of the noc node can be changed.
> > > > >
> > > > > Fixes: ad1abc8a03fdbc05b ("arm64: dts: imx8mq: Add interconnect
> > > > > for
> > > > > lcdif")
> > > >
> > > > Martin,
> > > >
> > > > Do you have any comment?? So your commit added something
> > > > untested?
> > > >
> > > > Shawn
> > >
> > > hi Shawn,
> > >
> > > well, for imx8mq the only missing piece is the mxsfb icc bandwidth
> > > request. I posted a first version a year ago but that didn't make
> > > it
> > > in:
> > > https://lore.kernel.org/linux-arm-kernel/[email protected]/
> > >
> > > So this should create a working state until the real fix in mxsfb
> > > is
> > > there (although I'd revert commit ad1abc8a03fd ("arm64: dts:
> > > imx8mq:
> > > Add interconnect for lcdif") instead).
> >
> > Besides lcdif, the mipi-csi devices have interconnects property too.?
> > Are
> > they already working?? If so, it makes more sense to revert
> > ad1abc8a03fd
> > instead.
>
> imx8mq-mipi-csi.c request the bandwidth, yes. we use that and the
> preliminary mxsfb request above so that works.
>
> do you want me to send the revert?
Yes, please.
Shawn